Video Encoder Transform Selection for Windowed Pattern Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards, such as the MPEG-4 AVC Standard, lack transform selection flexibility, leading to suboptimal compression performance due to the use of a fixed discrete cosine transform (DCT) for each block size, which can result in artifacts like the 'windowed pattern' at low bit-rates.
Innovation Solution
The implementation of a video encoder and decoder that selects between a first transform, typically the DCT, and a second transform, such as the Karhunen-Loeve Transform (KLT), with the second transform's coefficients determined responsive to a deblocking filter to avoid the 'windowed pattern' artifact, by ensuring the lowest frequency component is smooth to be effectively filtered out.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed DCT transform is used for each block size, then the coding standard is simple and easy to implement, but compression performance is suboptimal and artifacts like 'windowed pattern' occur at low bit-rates
Solution Approach 1:
The patent introduces dynamic transform selection where the transform type (DST or DCT) is chosen adaptively for each block based on local characteristics such as gradient magnitude and direction. This allows the coding system to transition from a static fixed transform to a dynamic adaptive transform, improving compression performance while maintaining implementation feasibility through simple selection criteria.
Solution Approach 2:
The patent changes the transform parameter (type selection between DST and DCT) based on local image characteristics. By computing gradient metrics and selecting transforms accordingly, the system adapts the transform parameter to match local content properties, thereby improving compression efficiency and reducing artifacts without significantly complicating the coding standard.
2Manufacturing precision
If transform selection is introduced to improve compression performance, then artifact reduction is achieved, but device complexity increases
Solution Approach 1:
The patent divides the image into blocks and applies different transform selection strategies to different blocks based on their local characteristics. Each block is independently evaluated using gradient metrics, allowing localized adaptivity without requiring global optimization. This segmentation approach limits the complexity increase to individual block processing rather than entire frame processing.
Solution Approach 2:
The patent implements partial transform selection by applying the more complex DST transform only where necessary (based on gradient criteria) rather than universally. This selective application achieves artifact reduction in problematic areas while maintaining simplicity in areas where DCT suffices, thereby limiting the overall complexity increase.
3Productivity
If multiple transforms are used with selection, then compression efficiency improves, but the number of transforms is limited in prior art
Solution Approach 1:
The patent makes the transform selection mechanism universal by applying the same selection criteria and transform pair (DST/DCT) across different block sizes and content types. The gradient-based selection approach works consistently regardless of block dimensions or specific content characteristics, providing a versatile solution that improves compression efficiency without requiring multiple specialized transform sets.
Data Source
AI summary
Methods and apparatus are provided for constrained transforms for video coding and decoding having transform selection. An apparatus includes a video encoder for encoding at least a block in a picture by selecting a particular transform from at least a first transform and a second transform that is different than the first transform, and applying the particular transform to a residue of the block. Transform coefficients for the second transform are selected responsive to a deblocking filter that is present in the video encoder.